On 10/12/22 17:30, Thomas Monjalon wrote: > 07/10/2022 13:55, Ilya Maximets: >> On 10/7/22 13:50, Ilya Maximets wrote: >>> On 9/12/22 12:09, Thomas Monjalon wrote: >>>> 16/03/2022 13:01, Ilya Maximets: >>>>> 'has_vlan' attribute is only supported by sfc, mlx5 and cnxk. >>>>> Other drivers doesn't support it. Most of them (like i40e) just >>>>> ignore it silently. Some drivers (like mlx4) never had a full >>>>> support of the eth item even before introduction of 'has_vlan' >>>>> (mlx4 allows to match on the destination MAC only). >>>>> >>>>> Same for the 'has_more_vlan' flag of the vlan item. >>>>> >>>>> Changing the support level to 'partial' for all such drivers. >>>>> This doesn't solve the issue, but at least marks the problematic >>>>> drivers. >>>> >>>> You changed "eth" and "vlan" from "Y" to "P". >>>> The field "has_vlan" is part of "rte_flow_item_eth", >>>> and "has_more_vlan" is part of "rte_flow_item_vlan", >>>> so I agree we need to change both items to "partial support". >>>> It looks to be a good change, just needs to more explicit, >>>> adding this kind of explanation about the fields. >>> >>> I can add this to the commit message. Should I also add >>> some note alongside the table in that documentation page? > > I'm afraid it can be long and complex to add notes in the page > about what is missing to get complete support for each feature/PMD. > I think you can just re-spin with a clear explanation in the commit, > so PMD maintainers can refer to it.
OK. v2 sent here: https://patches.dpdk.org/project/dpdk/patch/20221013104849.2677995-1-i.maxim...@ovn.org/ Sorry, forgot about --in-reply-to. Can re-post if necessary. > >>>> We missed this patch in 22.07, let's have some progress quickly. >>> >>> It was a busy month + PTO, sorry I didn't get to that patch faster. > > No problem, we are still on time for 22.11. > >>>>> Some details are available in: >>>>> https://bugs.dpdk.org/show_bug.cgi?id=958 >>>> >>>> About rte_flow_item_eth.{src,dst}, I don't find a deprecation notice >>>> about it in the history of the file doc/guides/rel_notes/deprecation.rst >>> >>> It took some time to find the deprecation notice, but I did find it >>> in the end. It's in the doc/guides/rel_notes/deprecation.rst, and >>> it says: >>> >>> * ethdev: The flow API matching pattern structures, ``struct >>> rte_flow_item_*``, >>> should start with relevant protocol header. >>> Some matching pattern structures implements this by duplicating protocol >>> header >>> fields in the struct. To clarify the intention and to be sure protocol >>> header >>> is intact, will replace those fields with relevant protocol header struct. >>> In v21.02 both individual protocol header fields and the protocol header >>> struct >>> will be added as union, target is switch usage to the protocol header by >>> time. >>> In v21.11 LTS, protocol header fields will be cleaned and only protocol >>> header >>> struct will remain. > > You're right, this is the general notice for rte_flow_item_*. > >>>> This is what we have in the code: >>>> union { >>>> struct { >>>> /* >>>> * These fields are retained for compatibility. >>>> * Please switch to the new header field below. >>>> */ >>>> struct rte_ether_addr dst; /**< Destination MAC. */ >>>> struct rte_ether_addr src; /**< Source MAC. */ >>>> rte_be16_t type; /**< EtherType or TPID. */ >>>> }; >>>> struct rte_ether_hdr hdr; >>>> }; >>>> >>>> Do you think we should remove the old fields now? >>> >>> From the OVS perspective, we do not care much. OVS is still using >>> old fields, but it's fairly simple change that we can do while moving >>> to a new version of DPDK. >>> >>> From the perspective of the API clarity, it's probably better to clean >>> up these structures. >> >> There seems to be some outstanding work still though: >> >> https://patches.dpdk.org/project/dpdk/patch/20210312110745.31721-1-ivan.ma...@oktetlabs.ru/#129161 > > I think nobody took care of it so far. > I will send a patch to clean the easy ones, > and I will follow up with Ori for a complete plan. > > Thanks > >
